Real-time optimal path planning and wind estimation using Gaussian process regression for precision airdrop

This paper presents a time-critical cargo drop strategy that allows a fixed-wing unmanned aerial vehicle (UAV) carrying a cargo under an unknown wind field, to accomplish the cargo drop mission within the least amount of time while minimizing the cargo landing error. Specifically, we treat the spatial wind distribution as a noisy vector field and apply the Gaussian process (GP) regression method to estimate the wind model. In order to optimize the strategy, the objective function to be maximized has been chosen as the weighted sum of two conflicting objectives: more knowledge of the wind field and less travel time. We present some simulation results to compare the performance of the proposed strategy with a conventional method. Results demonstrate the advantage of the proposed method in terms of accuracy and multi-functionality over the non-estimation strategy.